« TOP Camera of the Year 2012 | Main | Can I Change My Choice? »

Friday, 04 January 2013


Feed You can follow this conversation by subscribing to the comment feed for this post.

Agree completely with your opinion of the Foveon sensor - similar to Kodachrome film. Yes there are limitations, but unsurpassed color rendition and sharpness. Great designs never die. Imagine a full frame Foveon-type sensor on a Leica (M-F ?)

The Bayer CFA will never be "as antiquated" as the Lumière Autochromes because the Autochrome was never popular and suffered a fundamental problem of not being reproducible. It's only digital reproduction that has brough the Autochromes back to life (in BBC documentaries and large back-lit transparency print).

I wouldn't rule out layered photodetectors but for them to work well (to have high quantum efficiency) you have to stop all the photons of a given bandpass in the layer. This is the problem Foeveon has. Foveon is closer to Autochrome in this regard.

An organic dye based layered systems seems like the best bet but even after a working device is made making this type of multilayered sensor in quantity is not full understood.

The Bayer CFA I think will be around for a long time yet.

....And those "old" cameras will become cult objects, just like today we have audiophiles who lust for vinyl LPs, and photogs who will only give up their Leica M2s loaded with Tri-X (or, well, maybe these days, Ilford HP5) when they're pried from their cold, dead hands.

Cameras are indeed much more interesting now, and both Panasonic and Olympus deserve the credits for having started the ball rolling when they introduced Micro 4/3. Crucially, Olympus sent the message that it was OK to make beautiful cameras again when they launched the E-P1. If it weren't for this brilliant little machine, mirrorless cameras would all probably look like DSLRs (like the Panasonic G1/GH1) or point-and-shoots on steroids like the insipid Canon EOS M.
This is, IMHO, the reason cameras have become so much more interesting than what we had five or so years ago. The new designs brought back pride in ownership, something that got lost in the way when all camera makers started making ergonomically oriented bodies. (I wouldn't go so far as to call them "melted soapbars", as some do.) Now we can have cameras that are a delight to behold, the Fujis being currently the best example of the kind.
Re. the Foveon sensor, I believe it nods to the future. I may be wrong, but it is so clever and effective (at least from what I see on the 'net) that most sensor makers will adopt this technology sooner or later. It's all up to Sony...

So far, the Foveon sensor is the Wankel engine of photography.

Not having to bear the pressures of a professional, and knowing that a film/digital hybrid existence is, in fact feasible, I can comfortably bask and luxuriate in the complete and utter obsolescence that my favorite classic film cameras now afford me.

I've thus far managed to evade the introduction and perpetual refinement of non layered sensors; think I'll skip the whole layered scenario as well- if they're still making film.

Foveon isn't the future. Dense bayer arrays are the future. You're already seeing it in cell phone cameras. Why try to snake photos unmolested down a deep well when you can just super sample.

On a slight tangent - aren't some of those autochromes just the most beautiful things you've ever seen?

And let's not forget that Fuji has taken the gamble on its X sensor bu moving away from the pure Bayer array pattern, obviating the need for the AA filter. Maybe future sensor design will take this path instead (one-layer sensors but in more useful patterns?).

I predict none of this will do anything to improve the standard of photography ;)

Let's hope you're right. Bayer array is a curse. I want 2013 to be the year of B&W sensor :)

On-sensor PDAF will make 2013 even better than 2012! The just-leaked Fuji X100S is supposed to have it. DMDs at last!

There is one limit on the effectiveness of the foveon type sensor and that is the spectral response of the silicon components. With a bayer array you have a large choice in the filters that can be put in the way of the colour blind silicon sensor. With foveon the sensor *is* the filter. I hope that silicon spectral response is good enough because what I've seen of foveon is incredible in terms of picture quality in other respects.

I doubt that layered sensors will be the future. They suffer from poor high ISO (and this is not about their current implementation, by design some channels will get less light then others) and they are probably much harder to manufacture. The real question is - what kind of advantage does such a sensor have over simple bayer pattern ? No AA ? Just make sensor with 4x more pixels and voila (bayer = simple = cheaper). Colors ? Come on, in PhotoShop era ? DR ? It's no problem now.
I really cannot see any real advantage. I would say that Bayer is like democracy and Foveon is more like Utopia - most people dream about it, but no one will ever implement it rigth.

I remember a couple of things from the past about the digital technology :
first, many years ago I read somewhere a report (maybe from MIT but not sure) saying that silver price increasing and resources diminishing had push towards a different kind of photo technology no more based on silver: it seemed to be science fiction!
second, more personal, again years ago, but not so many: chatting with a friend about digital photography I told him: look at the offer of major camera produce, they have 5-6 items based on film and only one or two digital, expensive and with many limitations (it was the era of the "fabolous 3-5 MP). It's clear that photography is film based, digital is only an experiment. When there will be more digital cameras than film ones I'll consider it :-). Times fly, technology changes...at my age I'm even curious to try an EVF based camera...
PS: only dilemma, OM-D or EX-1 ? Hmm, decisions....

I can make film look like an autochrome, but not digital without spending all day trying to do it...I say bring bring back autochrome! It's all been downhill since then, and I'm only half joking...

The relative success or lack thereof Foveon-style sensors is an interesting case study in the realities of engineering.

The Foveon chip is ultimately a tradeoff between spatial resolution and color resolution. What the Foveon guys are trying to tell you is "we can capture the 'real' color at each pixel!". So, their color resolution is in theory better than Bayer sensors, but they give up a lot on things like pixel density since they are doing their own thing and can't leverage what Sony, Canon or Nikon know about how to put 36 million wells on a chip.

But the reality is that it seems to me that capturing photons at multiple layers in a detector and turning that into color information is not all that much easier than doing Bayer reconstruction. And, Foveon is also missing the thousands of man hours that Sony, Canon and Nikon have poured into making Bayer color reconstruction really really good. And let's not even talk about the recent jumps in noise performance.

I'll believe Bayer sensors are dead when someone ships a D4 using something else. But I don't see it happening. Bayer color is certainly good enough.

Meanwhile, I'm happy enough to start carrying around my little Olympus E-PL5 which is barely bigger than a P&S is in many ways better than my D200 from four years ago.

Yes, just the fact alone that Sony brought us *two* cameras setting new records in compactness per sensor size is telling about how miniaturisation is going.

And I've been waiting for that.
Maybe it'll go farther yet, we are still far from as compact as they could make a film 35mm camera.

I protest - as do top hat wearers everywhere....!

There is definitely something about that autochrome - a realism maybe, I don't know. The representation of the tablecloth caught my eye, for one thing - like Vermeer maybe - although I expect to be shot down in flames for that comparison!

Sony, Canon and Panasonic have all lodged patents for sensors that capture true RGB data at every photosite.

What seems very odd to me is that Sigma (and Foveon before it) has singularly failed to license the technology to a major camera maker. Can you imagine Sony's chip making capabilities and Foveon's technology? Or what about Canon with a .18 micrometre process and such technology?

Indeed, Canon's singular inability (or unwillingness?) to deliver latest semiconductor technology to its camera lines might lead one to believe they are doing something totally different - let's hope so.

This picture reminds me of the old logo featuring Nipper the Fox Terrier listening to his master's voice.

"I say bring bring back autochrome! It's all been downhill since then, and I'm only half joking..."

The prettiest print I possess is a 1910 platinum print, a portrait of a relative. Yet you know, inkjet prints can come pretty close. It's all been sideways from there, maybe?


*If* you could make Foeveon/multilayer system work as well (same sensel sensitivity and pixel color fidelity) as the best Bayer CFA on a single layer sensor you would gain a bit over a stop more sensitivity and about 1.4x the resolution plus avoid any CFA related color moiré issue. Not a lot of gain for quite a lot of development pain.

Until you can better the QE *and* color fidelity from a Bayer CFA on single layer of CMOS pixels multilayer pixels are not going to take off. I don't see that changing before Bayer and conventional single layer CIS maxes out perhaps in the next decade or so.

My favorite comment is Auntipode's "So far, the Foveon sensor is the Wankel engine of photography."

The Foeveon is also the Zeppelin of sensors. It really seems like "The Future" but you just can't get it to work well except in very special cases.

All that said one should never say never. My crystal ball isn't that good. Single later sensors will stagnate at some point. That's why all the sensor makers are looking at layered sensors and writing patents. Just in case.

RE Huw Morgan's comment: a camera with an EVF can never be that DMD, because the moment I see the action is the exact moment past that action.

Mike, yeah, you can get pretty close...maybe sideways...

I remember looking at big blow-ups of Farm Administration Kodachromes from the Library of Congress collection, back when I lived in DC, and I was "wowed" by the look of actually old Kodachrome through the scanning process. If any one remembers, Kodachrome was notoriously hard to "dupe", even into the 90's, and all the dupe material was Ektachrome and it just didn't "look right", or it was color negative and "sort of" close. But, boy, scanning really got that color nailed! So there's something to work with in the computer process...

Saw a back-lit display of Autochromes in San Francisco once, back in the mid-2000's, and it was stunning, tho.

As for Foveon vs. the world, to this day, everything I see out of that Sigma camera looks like film, and everything I see out of my cameras, is "sort of" with work. I wouldn't count out Foveon yet. I'd own one if I could afford it! Now if you could make a Foveon camera putting out .tiffs, we're on our way!

BTW, for Autochrome lovers, anyone remember Ansco 500 transparency? We used to use this back in the 70's, it was grainy as hell, even in 120, and had an unnaturally warm palette, but it actually looked like Autochromes...color faded fast, tho...

Remember, the human eye has higher luminance resolution than color resolution as well. The Foveon sensor, and optically equivalent designs, glories in bringing the two together, making them match -- at which point you have higher color resolution than your target market can use at any given spatial resolution.

Also, just a rhetoric alert -- color film is constructed in layers too, so it's not really "safe" to claim you're avoiding layered sensors by shooting color film!

With a 60Hz refresh the maximum lag is a frame period, 17mS. Faster refreshes the lag goes down. Some cameras are running at up to 240Hz refresh now for their AF (and OLED EVF?).

The shutter lag on a Leica (the ur-Decisive moment camera) is around 15mS. About the same as the 60Hz refresh lag.

Your simple reaction time is around 200mS. Recognition reaction times are closer to 400mS. With a warning you can speed up your reaction time. Even your brain's visual system has a lag of 20 to 40mS compared to the real world. I think you overestimate your speed of response compared to the equipment.

A modern EVF is not a limit to the DMD camera.

@keru: I want 2013 to be the year of B&W sensor

What I'd like to see is a monochrome sensor that's designed for exposure range. Instead of just leaving off the filter array, put on one that uses neutral density filters instead of colors. A simple implementation would be a checkerboard array with "white" squares unfiltered and "black" squares with a strong ND filter. It would cost you some resolution and increase shadow noise a bit, but you'd gain highlight range roughly equal to the density of the ND filters. That seems like a worthwhile price to pay for a big extension of exposure range.

@Andreas: a camera with an EVF can never be that DMD, because the moment I see the action is the exact moment past that action.

That depends on the sophistication of the camera. If it has a fast reading sensor and a lot of memory, you could have it buffer several frames and grab one from before you pushed the shutter: effectively negative shutter lag. Or you could store the few frames before and after the shutter push, so you can choose the best one later. That sounds like just the thing for capturing the decisive moment.

The comments to this entry are closed.



Blog powered by Typepad
Member since 06/2007